Relationship between electricity demand and temperature: methodology for load forecasting using weather numerical simulation

This work aims to study the relationship between the demand for electricity and the temperature for the concession area of a given electricity distributor, in order to develop a representative load forecast method for this area. For this purpose, a weather numerical simulation is made using the regional meteorological model Weather Research and Forecasting (WRF) for the study area and, from that, the average temperature in the distributor's concession area is calculated, considering the number of consumers in each city that makes up that same area, called Average Temperature Weighted by the Number of Consumers (TMPNC). With this, a temperature index is obtained that effectively describes the influence of temperature on the load, since a greater weight is given to temperatures in cities where there are more consumers. Thus, it is observed that the relationship between the load and the TMPNC differs depending on the month, day of the week and time of day, due to different causes related to human activity, socioeconomic factors, weather conditions, etc. However, a load forecast method is developed based on a regression algorithm that uses the load and TMPNC data as input, returning regression functions that describe the load behavior as a function of the TMPNC. The forecasting method developed resulted in load estimates with an mean absolute percentage error of 1.8%.